What does this skill do, and when should you use it?
Ad Creative Builder is the paid-ads creative production skill in Aaron Marketing Skills’ ROAS workflow. It produces RSA headlines and descriptions, hooks, and an angle matrix, with each unit tied back to a real claim on the destination page. It uses the offer, platform, audience, brand voice, narrative canon, and claims projection to guide drafting. It does not audit an ad account or optimize the post-click landing page.
Reads the offer, destination URL, platform and format, audience or intent, existing variants, and named offsets in memory/projections/narrative.json and claims.json; extracts the destination page headline, value proposition, concrete offer or claim, and CTA; builds three or more angles with hooks and ad variants; writes RSA headlines, descriptions, and hooks within the target format limits; annotates each unit with the destination claim it echoes; checks message match, unsupported claims, superlatives, guarantees, and policy risks; and produces the standard handoff summary. Unresolved claims may be submitted through registry-events.py as authorized proposals.
- A paid-search team needs a batch of RSA headlines and descriptions for a product or offer.
- A social advertising team wants multiple hooks and creative angles for testing.
- A marketer has losing headlines and wants to keep winners while replacing the rest.
- A brand team needs ad copy to echo the destination page’s headline, value proposition, and CTA.
How do you install this skill?
- Do not publish or import the generated set directly into an ad account; verify current platform limits, actual landing-page claims, claims provenance, and restricted-category policies first.
- The shared scripts, claims/narrative projections, and referenced materials are not included in the supplied evidence. Treat missing dependencies as blocked or NEEDS_INPUT rather than assuming dependency_status is verified.
- SECURITY.md lists 17.x as the current supported line while the skill and README use 18.0.0; confirm supported-version and maintenance status before adoption.
- For Meta, Google, TikTok transparency libraries, or destination-page fetching, confirm mainland-China reachability, authorization, robots/TOS compliance, and data-egress implications.
- Shell / CLI
- Network access
- Local filesystem
The skill is located at ad/orchestrate/ad-creative-builder/SKILL.md. To install the collection, Claude Code users can run /plugin marketplace add aaron-he-zhu/aaron-marketing-skills and then /plugin install aaron-marketing@aaron. Other Agent Skills-compatible hosts can use npx skills add aaron-he-zhu/aaron-marketing-skills. The README does not document a generic single-skill installation command.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/aaron-he-zhu/aaron-marketing-skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/ad/orchestrate/ad-creative-builder" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Provide the product or offer, destination URL, platform and format, audience or intent, brand voice, and ROAS profile. Example: Generate 15 RSA headlines and 4 descriptions for [product/offer], destination [URL]. Without a destination URL, the skill cannot enforce message match and should follow the NEEDS_INPUT path.
What are this skill's strengths and limitations?
- Produces concrete, ready-to-import creative units rather than only high-level advice.
- Enforces per-unit destination-page message matching and requires at least two angles.
- Includes format-limit, claims, and policy pre-checks.
- Can use first-party account exports and public ad-transparency libraries for iteration.
- Works at Tier 1 without requiring paid advertising APIs.
- A destination URL is required for complete message matching; missing narrative or claims evidence prevents a publish-ready result.
- The workflow depends on repository reference materials for format specifications, angle patterns, humanizer checks, and shared handoff rules.
- The source does not provide platform-specific test coverage; Google Ads and Meta APIs are optional rather than required.
- When evidence is missing, it preserves [needs source] and marks the dependency blocked instead of inventing support.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Ad Creative Builder this page | 60 · Recommended | ★ 3k | 12d ago | Apache-2.0 |
| Paid Ads Conversion Signal QA | 56 · Use with care | ★ 3k | 12d ago | Apache-2.0 |
| Ad Landing Experience Checker | 55 · Use with care | ★ 3k | 12d ago | Apache-2.0 |
| Ad Creative & Copy Generation Skill | 49 · Use with care | ★ 2.8k | 7mo ago | MIT |
| Email Creative Builder | 58 · Recommended | ★ 3k | 12d ago | Apache-2.0 |
This skill creates ad creative. ad-account-auditor scores the creative and account against ROAS, landing-optimizer handles the post-click page, and content-writer handles organic articles.
How did FollowSkills review this skill?
The skill treats exports, scraped pages, and competitor ads as untrusted data, requires input confirmation, saves only after user confirmation, submits unresolved claims as propose events, and explicitly disallows account upload or activation. These controls reduce overreach and prompt-injection risk. Deducted 6 points because the referenced runtime, permission boundaries, rollback details, and complete data flows are not included in this skill evidence; reading the destination URL also entails external network access, while sensitive-data handling largely depends on shared documents.
Inputs, outputs, procedure, quality bar, and the NEEDS_INPUT path are clearly specified, and format-limit checks support consistent behavior. Deducted 11 points because static review cannot reproduce key paths; the referenced scripts and reference materials are not supplied here, platform specifications are explicitly subject to change, and version 18.0.0 conflicts somewhat with SECURITY.md identifying 17.x as the current supported line. The static ceiling keeps this at 9/20.
Triggers, argument hints, platforms, required inputs, output shape, and neighboring non-fit skills are clearly defined; the core path can use user-provided data and includes Chinese descriptions. Deducted 4 points because platforms, ad-transparency libraries, and destination URLs require external network access, mainland-China reachability and language/platform variance are not sufficiently documented, and non-Google/Meta format coverage is relatively coarse.
The skill has structured frontmatter, Quick Start examples, a Skill Contract, data-source guidance, numbered instructions, save-template references, next-skill routing, and an Apache-2.0 license. The repository README adds versioning, architecture, CI, and maintenance signals. Deducted 6 points because the shared protocols, scripts, and referenced materials are not expanded in the supplied evidence; there is no skill-level changelog, clearly assigned maintainer, or troubleshooting FAQ, and the support-version statement is not fully aligned with the skill version.
The stated output is a ready-to-import creative set containing RSA headlines, descriptions, hooks, an angle matrix, and per-unit message-match notes. Character limits, angle coverage, claim checks, and policy pre-checks define a useful quality bar. Deducted 8 points because no representative output or third-party execution evidence is supplied, and direct importability depends on current platform specifications, claims projections, narrative canon, and referenced scripts that are not included. The static ceiling limits this to 7/15.
The skill requires destination-claim mapping, provenance labels, dependency tuples, and explicit unresolved-claim states; repository CI, regression tests, and an audit-artifact validator provide some auditability. Deducted 5 points because the supplied tests primarily cover architecture and generic audit artifacts rather than this skill's key generation paths, and no third-party execution results or independent reproduction evidence are provided. The static ceiling limits this to 5/10.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →